JujubeImageBD: A Comphrehensive Image Dataset of Jujubes Varieties in Bangladesh for Identification and Classification Using Machine Learning and Computer Vision
收藏NIAID Data Ecosystem2026-05-02 收录
下载链接:
https://data.mendeley.com/datasets/dydrxfpxm7
下载链接
链接失效反馈官方服务:
资源简介:
Type of data: 1024 x 1024 px images of jujubes.
Data format: JEPG.
Dataset contents: Original images of different varieties of jujubes in Bangladesh from single-fruit perspective.
Number of classes: Five jujube varieties - (1) Apple Kul, (2) Ball Sundari Kul, (3) Bau Kul, (4) Deshi Kul, and (5) Narkeli Kul.
Total number of images in the dataset: (A) In Original Dataset = 1,716 images and (B) In Augmented Dataset = 17,160 images.
Distribution of instances:
(A) Number of images in Original Dataset (Jujube Dataset):
(1) Apple Kul = 289.
(2) Ball Sundari Kul = 303.
(3) Bau Kul = 356.
(4) Deshi Kul = 356.
(5) Narkeli Kul = 412.
(B) Number of images in Augmented Dataset (Augmented Jujube Dataset):
(1) Apple Kul = 2,890.
(2) Ball Sundari Kul = 3,030.
(3) Bau Kul = 3,560.
(4) Deshi Kul = 3,560.
(5) Narkeli Kul = 4,120.
Dataset size: (A) Total size of the Original Dataset is 41.7 MB and the compressed ZIP file size is 39.6 MB. (B) Total size of the Augmented Dataset is 1.2 GB and the compressed ZIP file size is 1.11 GB.
Data acquisition process: Images of jujube are captured using a high-definition smartphone camera from different angles.
Data source location: Local wholesale and retail fruit markets located in different areas of Dhaka, Bangladesh.
Where applicable: Training and evaluating machine learning and deep learning models to distinguish jujube varieties in Bangladesh to support automated identification and classification systems of various jujube varieties which can be utilized in areas of computer vision, smart agriculture and horticulture, precision farming, supply chain automation, food industry, AI-based fruit recognition, ecology and ecosystem health monitoring, biodiversity efforts, botanical research, environmental conservation, educational resources, ecology research.
数据类型:分辨率为1024×1024像素的枣果图像。
数据格式:JPEG(原文笔误标注为JEPG)。
数据集内容:孟加拉国不同品种枣果的单果视角原始图像。
类别数量:共5个枣品种,分别为(1) 苹果库尔(Apple Kul)、(2) 球桑达里库尔(Ball Sundari Kul)、(3) 鲍库尔(Bau Kul)、(4) 德西库尔(Deshi Kul)、(5) 纳克利库尔(Narkeli Kul)。
数据集总图像量:(A) 原始数据集:1716张;(B) 增强数据集:17160张。
样本分布:
(A) 原始数据集(枣果数据集)各品类图像数量:
(1) 苹果库尔(Apple Kul):289张;
(2) 球桑达里库尔(Ball Sundari Kul):303张;
(3) 鲍库尔(Bau Kul):356张;
(4) 德西库尔(Deshi Kul):356张;
(5) 纳克利库尔(Narkeli Kul):412张。
(B) 增强数据集(增强版枣果数据集)各品类图像数量:
(1) 苹果库尔(Apple Kul):2890张;
(2) 球桑达里库尔(Ball Sundari Kul):3030张;
(3) 鲍库尔(Bau Kul):3560张;
(4) 德西库尔(Deshi Kul):3560张;
(5) 纳克利库尔(Narkeli Kul):4120张。
数据集大小:(A) 原始数据集总容量为41.7 MB,压缩ZIP包大小为39.6 MB;(B) 增强数据集总容量为1.2 GB,压缩ZIP包大小为1.11 GB。
数据采集流程:采用高清智能手机从不同角度拍摄枣果图像。
数据采集地点:孟加拉国达卡市不同区域的本地果蔬批发与零售市场。
适用场景:可用于训练与评估机器学习、深度学习模型,以区分孟加拉国的枣果品种,为各类枣果品种的自动识别与分类系统提供支撑;可应用于计算机视觉、智慧农业与园艺、精准农业、供应链自动化、食品工业、基于AI的果蔬识别、生态与系统健康监测、生物多样性保护、植物学研究、环境保护、教育资源开发以及生态学研究等领域。
创建时间:
2025-03-13



